IEEE Consumer Electronics Magazine - January/February 2023 - 111

Table 1. Types of attacks and their descriptions.
Serial
number
1
2
3
4
5
6
7
8
Attack type
DoS, DDoS
attack
Sniffing
attack
Brute force
attack
Integrity
attack
Web attack
Malware
attack
Fuzzy Attack
Description
Occupying and controlling nodes or network resources by sending a lot of requests or
unrelated data
Such as port scanning attacks, the goal of which is to steal confidential data of the
vehicle system
Cracking information such as the password of the vehicle network or system
Spoofing attack, replay attack, wormhole attack, Sybil attack, etc.
To invade the web pages of vehicles or node servers, attackers implement web attacks
through SQL injection or cross site scripting (XSS)
Attackers inject into the system through vulnerabilities in the communication interface
to cause attacks by the forms of the worm, viruses, spyware, etc.
Attackers force the vehicle into an unexpected state or malfunction by injecting
random information
Other types of attacks
Previously, research on intrusion detection for
IoV showed a fragmented state.4 Some researchers
focus on intrusion detection research for IVN,2,7
the CAN protocol does not provide security mechanisms
such as message authentication or data
encryption at the design stage, so for IVN security
researchers, the CAN bus is the main object of
research for IVN intrusion detection. Others focus
on out-of-vehicle networks.8-11 In this article, we
will not conduct a separateness study, but conduct
a comprehensive discussion of intrusion detection
researchon internal and external networks.
Considering the current research directions,
the IoV intrusion detection technologies are classified
as follows: misuse-based intrusion detection,
anomaly-based intrusion detection, and based on
hybrid or other intrusion detection.12 Compared
with the misuse-based method or other methods,
the anomaly-based method is the most promising
in the IoV intrusion detection field.13 Anomalybased
methods will be able to monitor known and
unknown attacks effectively, and artificial intelligence
(AI) technologies such as machine learning
(ML) are generally applicable to the problem of
abnormal intrusion detection in IoV.
AI-BASED IDS FOR IOV
At present, a considerable part of the existing
IoV intrusion detection models are based on AI.
January/February 2023
Compared with IDS based on non-AI, the AIbased
IDS (such as ML-based IDS) has the advantage
that it can automatically mine the specific
models in the data that cannot be mined manually,
and provides more accurate detection
results. Previous studies, however, have only
included AI-based IDS as part of the various
approaches they have reviewed. This article
goes beyond these previous works, and focuses
specifically on the intersection between AI technologies,
IIoV security threats, and intrusion
detection research. We cover a range of increasingly
important AI technologies for IoV intrusion
detection; these technologies have not been discussed
independently in previous studies. Table 2
provides a comparative summary of contemporary
typical AI-based IDSs for the IoV. We take the
detection performance, low latency, lightweight,
and generalization ability as thebasic performance
that the AI-based IDS for IoV needs to meet. The
proposed solutions may be deployed in the ECUs
of a vehicle, the RSUs of a VANET,14
deployment points. There are different IDS deployment
strategies for different detection tasks.
Traditional ML methods: Many traditional ML
methods have their unique advantages in feature
extraction and classification. Alshammari et al.13
clustered and further classified intrusion events
in VANET through K-nearest neighbor and support
vector machine algorithms. Yang et al.4
111
or other
IEEE Consumer Electronics Magazine - January/February 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2023
